Nvidia unveils RTX Spark superchip

Nvidia used Computex in Taipei this week to push a new argument about where artificial intelligence computing goes next. On June 1, Chief Executive Jensen Huang unveiled the RTX Spark Superchip, a PC platform that combines an Arm-based CPU, a Blackwell GPU and as much as 128GB of unified memory for laptops and desktops. Nvidia said the systems are designed to run AI agents locally on Windows PCs rather than sending every task back to the cloud. On June 2, Huang added that Nvidia has enough supply to support strong CPU and GPU growth despite broader constraints in the semiconductor market. ### What exactly did Nvidia launch in Taipei? Computex opened with Huang presenting RTX Spark as a new PC platform rather than a single add-in graphics product. Nvidia’s own materials describe it as a Windows-on-Arm system built around an Arm CPU, a Blackwell-based GPU and unified memory, with the top configuration reaching 128GB. (nvidia.com) Tom’s Hardware reported the flagship configuration includes as many as 20 Arm CPU cores, 6,144 CUDA cores and up to 300 GB/s of memory bandwidth. Nvidia has also published a Spark roadmap, signaling that the company intends the launch as the start of a broader PC chip family. ### Why is Nvidia talking about “AI agents” on PCs? (nvidia.com) Nvidia said RTX Spark is meant to run AI agent software on-device, using local compute and memory instead of relying entirely on remote data centers. TechCrunch reported the company is pitching the systems as a way to bring AI agents to mass-market PCs from partners including Microsoft, Dell and HP. (tech.yahoo.com) The same report said Nvidia is targeting what it described as a $200 billion CPU market. That places the launch beyond Nvidia’s core data-center business and into a part of the PC stack long dominated by Intel, AMD and, more recently in Windows-on-Arm systems, Qualcomm. ### Which companies are signed up to ship machines? (techcrunch.com) More than 30 laptop makers and 10 desktop makers plan systems based on RTX Spark this autumn, according to reports from Computex. TechCrunch said initial Windows PCs using the chip are expected from Asus, Dell, HP, Lenovo, Microsoft Surface and MSI, with Acer and Gigabyte to follow. (techcrunch.com) Microsoft’s inclusion matters because Nvidia is tying the product directly to the future of Windows PCs. Dell and HP give the launch immediate reach in commercial notebooks and desktops if the announced systems arrive on schedule. ### How does this fit with Nvidia’s supply story? (finance.yahoo.com) Reuters reported from Taipei on June 2 that Huang said Nvidia has enough supply to accommodate robust growth in both CPUs and GPUs. He made the comment as investors continue to watch whether AI demand can spread from hyperscale data centers into PCs and other devices without creating another bottleneck. (techcrunch.com) CNBC reported analysts see the PC push as part of Huang’s effort to expand Nvidia’s reach across more of the AI stack. Nvidia has not said in the material reviewed here how many RTX Spark units it expects to ship this year, but Huang’s supply comments suggest the company is trying to reassure customers before autumn launches. (whbl.com) ### What should readers watch next? Autumn is the next concrete milestone because that is when Nvidia’s hardware partners say RTX Spark systems are due to reach the market. The first wave is expected to include machines from Microsoft Surface, Dell, HP, Lenovo, Asus and MSI, with Acer and Gigabyte following later, according to TechCrunch’s account of the launch. (finance.yahoo.com) (cnbc.com)